Graduate Thesis Or Dissertation
 

Applying Machine Learning Algorithms to Identify Problematic Nuclear Data within Nuclear Transport Simulations

Public Deposited

Downloadable Content

Download PDF
https://ir.library.oregonstate.edu/concern/graduate_thesis_or_dissertations/ft848x94q

Descriptions

Attribute NameValues
Creator
Abstract
  • The primary goal of this dissertation is to improve the quality of nuclear data available to the nuclear science community. We propose to accomplish this by applying machine learning algorithms to the large number of available benchmark experiments and simulations, with the goal of determining which nuclear data have strong relationships with simulation error. Combining these insights with expert knowledge can both inform new experimental designs and identify areas of nuclear data that should be reevaluated in order to efficiently reduce simulation error. Specifically, we use a random forest to predict the error of the simulated multiplication factor in benchmarks from the Whisper criticality benchmark suite, and then employ model explanation methods to understand the relationships between the error and associated nuclear data. We demonstrate the utility of this approach by identifying inconsistencies in the ¹⁹F inelastic scattering nuclear data. We then go on to show how the machine learning insights can be incorporated into the optimization of nuclear data to both target specific subsets in an intelligent manner and also speed up a stochastic optimization algorithm. This approach differs from past methodologies in that previous methods either adjust the entire nuclear data library, or rely on expert judgment to identify regions of nuclear data to modify. Both are sub-optimal: one adjusts data regardless of if it is truly influencing error while the other relies on human judgment which is sensitive to human bias. This approach does not rely on humans, is reproducible, and has the capability to both improve our understanding of nuclear data and the fidelity of our simulations. Finally this methodology is not limited to criticality benchmarks, but can be applied to fusion, activation, detection, and reactor benchmarks.
Contributor
License
Resource Type
Date Issued
Degree Level
Degree Name
Degree Field
Degree Grantor
Commencement Year
Advisor
Committee Member
Academic Affiliation
Rights Statement
Publisher
Peer Reviewed
Language
Embargo reason
  • Pending Publication
Embargo date range
  • 2022-01-01 to 2023-02-01

Relationships

Parents:

This work has no parents.

In Collection:

Items

Thumbnail Title Date Uploaded Visibility Actions
file details GrechanukPavel2022.pdf 2021-12-31 Public Download